Building high-quality prompts from scratch is time-consuming and can lead to inconsistent results. Custom prompt templates in Model Studio let you create structured, reusable templates. Use them to design, manage, and optimize prompts for stable, high-quality output.
This topic is applicable only to the Chinese Mainland Edition (Beijing region).
Create a text generation template
Use these templates for tasks like content creation, knowledge-based Q&A, code generation, and conversational interactions. They enable structured design through a prompt engineering framework.
Two Input Mode are supported:
-
Custom creation: Use this mode to quickly convert an existing prompt into a template or make simple optimizations.
-
Create based on prompt framework: Use this mode with a professional framework to build high-quality prompts for complex tasks, structure your ideas, and ensure the prompt's integrity.
Method 1: Custom creation
-
Go to the Prompts page, click Create Prompt, and configure the following parameters.
-
Template Name: Enter a custom name.
-
Basic Information: Select Text Generation.
-
Input Mode: Select Custom Creation.

-
In the text box, enter the prompt and click Optimize Prompt.
Example: Enter
Please help me write a 100-word sales data analysis report.Before optimization
After optimization


-
-
After optimization, click Save to create the template.

Method 2: Create with a prompt framework
-
Go to the Prompts page, click Create Prompt, and configure the following parameters:
-
Template Name: Enter a custom name.
-
Basic Information: Select Text Generation.
-
Input Mode: Select Create Based on Prompt Framework.
-
-
Select a Prompt Optimization Method. Different frameworks suit tasks of varying complexity. See Prompt engineering frameworks.
-
Fill in the corresponding fields based on the structure of the selected framework.

-
Click Optimize Prompt. The system refines and enhances your input.
|
Before optimization |
After optimization |
|
|
|
-
After optimization, click Save to create the template.

Create an image generation template
These templates are for image creation. You can define positive and negative prompts to control the content and style of the generated image.
-
Go to the Prompts page, click Create Prompt, and configure the following parameters:
-
Template Name: Enter a custom name.
-
Basic Information: Select Image Generation.
-
-
Enter the Positive Prompt and Negative Prompt.
A positive prompt tells the model what the generated image should include, while a negative prompt tells it what to exclude. Using both helps you generate images that meet your exact requirements.
Example:
-
Positive Prompt: Generate a brightly colored night sky with twinkling stars and a meteor streaking across. In the background is a tranquil lake, reflecting the glow of the stars. The scene should feel peaceful and dreamlike.
-
Negative Prompt: Do not generate any dark or blurry backgrounds. Avoid black or dark gray. Do not include city buildings, withered trees, or any oppressive elements. Keep the overall mood bright and pleasant.

-
-
After you finish, click Save to create the template.

Prompt engineering frameworks
The platform provides several built-in prompt engineering frameworks to help you structure complex tasks.
|
Framework |
Use cases |
|
ICIO |
Straightforward tasks such as data analysis, content generation, and text summarization. |
|
CRISPE |
Role-based interactions such as intelligent customer support, creative writing, and interview simulations. |
|
RASCEF |
Multi-step business workflows such as project planning, strategic analysis, and process design. |
Next steps
-
To create a custom prompt template with an API, see Create a custom prompt template by using an API.
-
To view, manage, and use custom prompt templates, see Prompt template overview.


